FW Ontology: classes and properties for fuzzy and wobbly data

An object representing a set of vague and uncertain values or ranges. The provided values or ranges represent the current best values. It might be further qualified by the hasConfidence property.

• fw:FuzzinessWobblinessResourceValue ^c
• fw:UncertainValue ^c
• fw:VagueValue ^c

Represents the most likely value, although it may also be a different value. It holds exactly one value; several candidate values are a fw:VagueValue or a fw:FuzzinessWobblinessValue.

Valuable object with unclear but limited references or ranges.

What a degree or another fuzziness value is based on, as found in the source.

• text evidence ^c
• media evidence ^c
• embedding evidence ^c

A text as given in the source, e.g. the string 'Lezoux or Vichy ?'.

An image, scan, audio or other media object, given by a link; e.g. the image an AI detection was run on.

A semantic embedding (vector) of a text, image or other object, given by a link to the vector and the model that produced it.

• has embedding ^op
• embedding model ^dp

Kind of fuzziness/wobbliness. The terms are individuals, so that fw:hasType points to a term, not to a class. Type and value class are two different things: the type (fw:Vagueness) says what kind of fuzziness a statement has; the value class (fw:VagueValue) is the container in which a value is stored. Each value class points to its type with rdfs:seeAlso. When a statement is derived from several others, AMT.engine gives it the common type of the antecedents; fw:Certainty does not count as long as another type is present, and different types give fw:FuzzinessWobbliness.

A procedure that produces or combines degrees of connection, confidences or other fuzziness values. Concrete methods are individuals and are resolvable by their IRI.

• formula ^dp
• implemented by ^op

• assignment method ^c
• propagation method ^c

Obtains a degree for a single statement from its source.

• rule-based method ^c
• statistical method ^c
• machine learning method ^c
• expert assessment ^c

Combines the degrees along a chain of statements into the degree of a derived statement.

A fixed rule applied to the form of the source statement.

A degree computed from observed data by a statistical procedure.

• correlation method ^c
• probability method ^c

A degree taken from the strength of a correlation between two series, e.g. Pearson.

A degree taken from a probability, classical or estimated.

A degree reported by a trained model, e.g. a classifier confidence.

A degree set by a person from judgement.

To indicate a quantitative statement about the certainty of a given statement. Could for example be used with: https://fuzzy-sl.wikibase.cloud/wiki/Main_Page - Prefix: fuzzy-sl - URL: https://fuzzy-sl.wikibase.cloud/entity/

The method by which a degree (or another fuzziness or uncertainty value) was obtained.

The kind of fuzziness/wobbliness a statement or value is affected by: vagueness, uncertainty, or the two together; or fw:Certainty for a statement that is not affected by any of them. The value is a term (an individual of fw:FW_Type), not a value class such as fw:VagueValue. A statement without fw:hasType is read as fw:Vagueness by AMT.engine (also at degree 1.0); a tool that wants to say that a statement is certain writes fw:Certainty together with a degree of 1.0.

Resolvable IRI of the code that implements the method (repository file, release or software DOI).

Optional. Links a statement to what its degree rests on: the evidence it was derived from (source string, image, embedding) or, for a derived statement, the statements it was derived from. Several statements may share one evidence node, which keeps them together (e.g. the alternatives of an OR).

Resolvable IRI of the media object (image, scan, audio).

Resolvable IRI of the vector (a file or an entry in a vector store). The vector itself is not written into the graph.

Indicates the confidence level [0–1] assigned by the contributor or the system. The closer the value is to 1, the higher the confidence. A value of 1.0 would correspond to a certain statement. A confidence is one kind of degree of connection (fw:hasDegreeOfConnection); the degree itself may describe something other than a confidence. Work in progress: how far the two coincide is still being discussed with the AFE data model.

• topDataProperty ^dp
• has degree of connection ^dp

For non resource values: numerical, years, ...

Degree [0-1] to which a connection (statement, edge) holds. Same meaning as amt:weight. The degree can describe anything that is measured between 0 and 1; fw:hasConfidence, an assessment by a contributor or a system, is a sub-property of it, not the other way round.

The calculation rule of a method as plain text, e.g. (1 / n_possibilities) * (1/2 if n_question else 1).

The text exactly as given in the source.

Name and version of the model that produced the embedding.